We studied Rhinicthys osculus and its close relatives to discover evolutionary processes that operated to produce this widespread, polytypic fish group in an intermountain landscape of hundreds of small, isolated drainages. This group has attracted study because of its many ambiguously distinctive populations, in which homoplastic traits are shared across local geographic barriers. The observed morphological ambiguity is clarified by phylogenetic analyses of mtDNA data from 73 locations, which show deep divergences separating several dozen hypothetically monophyletic groups. Calibration of genetic distances with fossil age estimates permits identification of three groups that originated in the late Miocene – centered in the Columbia-Snake, Colorado, and Lahontan drainage basins. These mtDNA groups contain Late Pliocene lineages within the Columbia and Snake rivers, upper Green River, Lower Colorado River, Humboldt River, Death Valley, and ancient connectives. Each of these clades accumulated local, differentiated populations by subdivision in Pleistocene internal drainages, but each also shows many populations transferred to neighboring basins by stream diversions, headwater captures, and pluvial lake overflows, which are consistent with geological evidence. Population groups defined by molecular analysis correspond only approximately to named or morphologically recognized taxa. Morphological analysis of 88 samples found no unique phenotypes. Three potential explanations (independently or in concert) may explain the evolution of the Rhinichthys osculus complex in the west: (1) Scores of drainage changes have alternated aquatic dispersal routes with long periods of isolation resulting in vicariant divergence. Drainage and fish transfers following long periods of isolation include, for example, the connection of the Upper to Lower Colorado River Basins (~5 Ma), the connection of the Snake River to the Columbia River (~3 Ma), many lake overflows among Pleistocene Lahontan basins documented by Marith Reheis and her colleagues, and the Lake Bonneville spillover to the Snake River and the Pacific (~15 Ka). Prior to such large connection events, basins and drainages were arranged differently and sometimes isolated for periods of 104 - 106 years or more, permitting accumulation of differences. (2) Alternating pluvial and arid stages associated with more than 20 glacial cycles and over 100 subcycles over the past 3 million years subjected populations to periodically intense selection for expansion in well-watered habitats, alternating with isolation in small populations in desert habitats. (3) Pluvial overflows created secondary contact and frequent genetic mixing, leaving only one form in the basin or stream; only two instances of sympatric species persist, both in large waters of the climatically more stable Columbia and Snake River drainages. Some traits that distinguish large-river species in the north also exist in southern desert populations, but with inconsistently mixed or convergent characteristics.